M********r
发帖数: 278 | 1 Although I don't time the market on large scale and keep my overall
allocation fixed, the gambler inside me, whom I will eventually kill off, do
allow myself about 5% room to do whatever i see fit. My equity allocation
has been 5% lower than target for more than a year now. The recent move in
10 yr yield is especially worrisome. |
|
t********t
发帖数: 5415 | 2 默认是taxable account?
不一定是“不少”，除非你大改allocation，但那样的话更需要担心的是新allocation
的盈利能力。 |
|
M********r
发帖数: 278 | 3 Expense ratio aside, this fund as well as other target date funds are
essentially index-based asset allocation models. You are not supposed to
judge them against a single equity index. The reason you pick this fund is
because it matches your risk tolerance. Once you have it, you should expect
it will for sure underperform an equity index such as S&P500 during bull
market. And that's not the reason to dump it.
However if you find its expense ratio significantly higher than other target
date funds... 阅读全帖 |
|
M********r
发帖数: 278 | 4 皇上固执归固执，但总心平气和的，这点我喜欢。 ok, 前面解释了tracking error,
自己去查查数据。我来解释7%的事。buy and hold a balanced portfolio, 长期下来
（〉十年）， 年化回报率6-7% 甚至更高（after fund ER, before tax on dividend
and share sales, nominal return)同样有据可查。最简单的办法你去查查vanguard
的那些balanced fund， 比如star fund, balanced index fund, wellington, 以及那
些asset allocation funds and target date fund.更不用说那些有名的asset
allocation models, 比如 coffee house, couch potato, Aronson family
http://www.marketwatch.com/lazyportfolio
我自己的60/40 portfolio (all VG funds, index an... 阅读全帖 |
|
M********r
发帖数: 278 | 5 I started investing in 2000. was in school. didn't have money. lived in a
shitty apartment in a shitty neighborhood. Everybody was talking about
stocks. Cisco, Juniper, Broadcom, JDSU, verisign, you name it. Oh yeh, pets.
com. I was probably born cautious. I somehow on the first day made up my
mind: I am not going to touch individual stocks. But the market was going up
everyday. couldn't resist. So I decided to buy MF. What was the biggest
name in MF business that time? Janus. The greek god with... 阅读全帖 |
|
s*********y
发帖数: 689 | 6 ER低的一般不会烂。这里头Vanguard Institutional Index应该是配置主力，加上
Vanguard Total Bond Admiral就是很不错的组合。具体的allocation比例取决你的
risk tolerance和投资的horizon。图省事当然也可以用target fund，它替你动态的调
整allocation和rebalance。
我总是建议读一点入门的投资方面的书或网站，为退休投资是重大的事情，需要有正确
的知识和方法。 |
|
w*****c
发帖数: 2130 | 7 刚刚开始工作，公司match一部分401k，不放就亏了，但又不懂投资，请大牛们帮忙看
下，指点一下，哪些可以买，比例如何，非常感谢，会有大包子奉上。
Investment Fund Expense Ratio
AB Global Risk Allocation Fund - Class A 1.31
Invesco Stable Asset Fund 1.10
PIMCO Real Return Fund - Class A 0.85
Invesco International Growth Fund - Class A 1.32
Goldman Sachs Growth and Income Strategy Portfolio - Class A 1.37
Fidelity Advisor Equity ... 阅读全帖 |
|
g***l
发帖数: 2753 | 8 这个东西有很多方法，取决于你自己的系统。
我前一阵就做过这个东西，几千个文件，方法就是就是查malloc, re-alloc,和free，
把这这三个函数从新包装，每次调用这三个函数的时间，能够打印出指针，分配或者释
放的内存的大小，还有相应的文件，行数信息，就是在什么地方调用这三个函数的。
运行程序直到memory耗尽死机，然后查log文件，就是如果一个指针由malloc分配出来的
，那么他就应该有相应的FREE去释放相同大小的memory，对于re-alloc也是一样的。
到最后，就会发现某处的指针是没有被释放的。
当然，到系统memory耗光的时间，有些内存是没有释放的，但是这些没有释放的内存你
可以分析他为什么没有被释放，当然memory leak就可以很容易的被找出来。 |
|
g*******y
发帖数: 1930 | 9 这些东西我很都不熟悉。希望有高手指点指点，呵呵
1. Mempool design with 30k limit.
mempool是应该在一开始就allocate 30k 连续的内存,然后分配和管理？
或者是每次call allocate(n)的时候再通过operator new[]来分配内存，update size
member？如果是的话，free(ptr, n)怎么写呢？貌似operator delete[]不能带size参
数啊？
总之我就是对memory design这块很不熟悉。。。
2. Implement put/get methods of a fixed size cache with LRU replacement
algorithm.
这个是不是用fixed size的max heap来实现？每个元素定义一个key，表示距离上次使
用的时间，每使用一个元素，就相当于是把它的key更新为比当前最小值更小的数，然
后做heapify()操作？
每put一个元素，就assign新元素一个最小的key，然后用新元素替换掉堆顶点，然后做
heapify?
3. Write |
|
a**x
发帖数: 154 | 10 We are given a singly-linked list A. The nodes of A consists of usual data
and next pointer. Along with these there is another pointer random which
points to a random node of A. We are supposed to make a new linked list B
which is copy of A as far as the structure is concerned but all its nodes
are newly allocated. Develop an algorithm such that its time complexity is O
(n).
似乎是小尾羊同学刚才的题目
我的问题就是这个linkedlist不可以直接一次loop，
for each node in original LL:
allocate new node
copy data
copy next poi |
|
g*******y
发帖数: 1930 | 11 i guess you can pre-allocate some memory in compile time, and then manage
that memory yourself (memory pool alloc/free), and then redefine the new
operator (class specific operator new) |
|
y**i
发帖数: 1112 | 12 如果把析构函数声明为私有的，那就是不能用delete了，那在heap上创建对象的话，如
何析构呢？首先要收回heap上allocate的内存的话，只能用placement new对吧？然后
还不能手动调用析构函数，必须要调用一个公用Destroy函数，在这个Destroy函数里调
用析构函数，最后再free在heap上allocate的内存，是这样么？ |
|
e******d
发帖数: 310 | 13 Sorry for no Chinese input.
I failed in Bloomberg on-site interview for a senior software development position. The following two questions may be helpful
1. In which cases can the destructor of a class not be called?
2. In a try block, we allocate some memory ; then an exception is thrown.
How can we get the allocated memory released before we jump to catch block to process the exception? |
|
s*****t
发帖数: 737 | 14 It seems to me that it depends on how operator+ is defined.
We can return a reference to a heap-allocated object and a local static.
Actually, we sometimes need to return a reference to a local static,
due to initialization order problem.
We can create a heap-allocated object, so it is not clear to me that a's
value has to be changed.
static,
value |
|
i**********e
发帖数: 1145 | 15 你这个答案不够优化，因为：
1）必须构造新的链表，需要额外空间。
2）必须把每一个元素给de-allocate，然后再allocate memory，需要额外的时间。
最优解是不需要变动元素的内存位置，变动的只是指针。
一些常见面试题的答案与总结 -
http://www.ihas1337code.com |
|
j*****g
发帖数: 223 | 16 1. all options are not mutually exclusive.
option a. overload new operator or try to shim malloc so that you can track
all the memory allocations, which line of code they originate, etc. Before
program exists, check allocation lists and see if any still left and what
file and what line caused the leak.
option b. use commercial leak detection program such as boundschecker (on
windows platform).
2. depending on OS. If on windows, you can override image load option in
registry key such as the OS wi... 阅读全帖 |
|
i**********e
发帖数: 1145 | 17 Your solution is basically a bottom-up approach. However, there is a problem
. You are allocating the vectors dynamically, which leaks memory after the
function finishes. How are you going to clean up? There is no easy way.
I know you allocate it dynamically instead of statically is to use pointer
to reference, thus avoiding copy by value for efficiency purposes. Another
problem is you are also using more memory than you should. When you found
the left branch has a greater depth, you don't clean... 阅读全帖 |
|
h*****1
发帖数: 74 | 18 来自主题: JobHunting版 - 问几道老题 In a try block, we allocate some memory ; then an exception is thrown.
How can we get the allocated memory released before we jump to catch block
to pr
ocess the exception? |
|
t**v
发帖数: 101 | 19 I think this question is asking how stl vector's implementation can be
modified. How about using a pointer to point to the newly allocated
memory? I.e. X->X, instead of allocating 2X and copying from X. |
|
c*********t
发帖数: 2921 | 20 写了一个很简单的程序，在linux下试了试,是segmentation fault, 程序退出。并且用
valgrind verify了一下，是同样的结果。
#include
// del_stack_variable.c
// this program tries to delete a variable in a function
//
int main()
{
int j=10;
printf("the addr of j =%p, j = %d\n", &j, j);
printf("delete j \n");
free(&j); //this will cause segmentation fault
printf("after that\n");
printf("the addr of j =%p, j = %d\n", &j, j);
return 0;
}
gcc -o del_stack del_stack_variable.c
$ ./del_stack
the addr of ... 阅读全帖 |
|
l********y
发帖数: 1327 | 21 i think this is ok but how do you allocate memory for dest * ?
i know in some compiler it will work like mine, but still dest * is not
allocated any memory. |
|
h*****g
发帖数: 312 | 22 1. In which cases can the destructor of a class not be called?
2. In a try block, we allocate some memory ; then an exception is thrown.
How can we get the allocated memory released before we jump to catch block
to process the exception? |
|
t******g
发帖数: 252 | 23 I don't understand how the second line in level one work. Can someone please
explain?
I think p is allocated as a pointer to pointer. But where it's pointed to is
not allocated. Why is that not a problem? |
|
d*******l
发帖数: 338 | 24 我不敢说完全明白，但有点感觉，谈谈我得想法。
就像上面说的，为了保证能返回一个对齐的地址且有足够的空间，需要打出一些富裕。
多少富裕足够呢？就是代码里的offset。我记得在OS课上老师提到过这种做法，好像叫
container，就是比实际需要多分配空间，把中间开始的某个地址返回，前面的地方存
放些信息。
下面说说offset怎么来的。假设我们用malloc分配了一段空间，起始地址是p1。此时我
们想得到一个对齐的地址，最直接的做法就是& ~(alignment-1)，这样低位清0，结果
就是对齐的。但这样做的结果相当于把原地址减去了一个[0, alignment-1]的数，也就
是低位被减掉了。得到的新地址的开始的一部分就是未分配的。代码中对p2进行这个操
作，得到的结果是p2=p1+offset-x，x就是在对齐过程中被减掉的数，可以看出p2>=p1+
1，这就保证了对齐之后的地址也是分配过的。
再说说p2[-1]=p1的作用。上面说的container方法有一个需要注意的就是free的时候，
如果直接free alloc返回的那个地址，其实是有问题的，因为那个是最初分配空间的中
... 阅读全帖 |
|
M*********t
发帖数: 257 | 25 here is what in vector header
explicit vector ( size_type n, const T& value= T(), const Allocator& =
Allocator() );
we only need default and copy constructor, that's all |
|
a***r
发帖数: 93 | 26 memory O(n)
time O(nlogn)
//~ Given a array,find out if there exist a subarray such its sum is zero
#include
#include
using namespace std;
static void swap(int *p, int i, int j) {
int t=p[i]; p[i]=p[j];p[j]=t;
}
static int partition(int *p,int left,int right) {
int j=left;
for(int i=left;i
if(p[i]
}
swap(p,j,right);
return j;
}
static void binary_sort(int *p, int left, int right) {
if(left>=r... 阅读全帖 |
|
d*******r
发帖数: 208 | 27 名单上列了11个人。见了10个,跟名单上有临时改动，有一个的shadow没来。
1. given such a structure,
id -- list of friend ids (sorted), similar to bi-directional or
undirected graph eg.
1 -- 3 5 6
2 -- 5 8
3 -- 1 8
5 -- 1 2
6 -- 1
8 -- 2 3 9
9 -- 8
find an efficient way to decide if two ids are connected by 1 hop or 2
hop. for instance.
one hop: 1 -> 3 -> 8
two hop: 1 -> 3 -> 8 -> 9
2. design a hashtable, get, put, etc. consider thread-safety and hash
capacity resizing.
3. some introduction of th... 阅读全帖 |
|
m**q
发帖数: 189 | 28 ==> 谢谢分享
名单上列了11个人。见了10个,跟名单上有临时改动，有一个的shadow没来。
1. given such a structure,
id -- list of friend ids (sorted), similar to bi-directional or
undirected graph eg.
1 -- 3 5 6
2 -- 5 8
3 -- 1 8
5 -- 1 2
6 -- 1
8 -- 2 3 9
9 -- 8
find an efficient way to decide if two ids are connected by 1 hop or 2
hop. for instance.
one hop: 1 -> 3 -> 8
two hop: 1 -> 3 -> 8 -> 9
==> BFS
2. design a hashtable, get, put, etc. consider thread-safety and hash
capacity resizing.
==> has... 阅读全帖 |
|
a*******r
发帖数: 122 | 29 Look it in another way, it looks like a prohlem about managing memory pages.
There is about 1k*1k*10, so if you create a table of 1k entries, each of
them with a link to a 10k bitmap, the bitmap memory won't get allocated if
no interger is removed in its 10k range, so for an average case like 100's
integers , there will be about 100's of 10k/8 memory get allocated, which
would give you an average case of less than 1M (up to 800 integers). |
|
i**********e
发帖数: 1145 | 30 是哪一道题呢？
我帮你看看吧。
你可以 return 空答案，然后看 large input 的testcase最大size，然后 allocate
那个size应该就好了。
面试时不用 dynamic alloc，assume input 不会超过这个size就好了。 |
|
l***i
发帖数: 1309 | 31 there needs to be some goals you want to achieve, e.g. you want allocation
to be fast, that is, in O(n) time or O(n^2) time you want to find a valid
rectangle for allocation.
Another goal is to be space efficient, that is, you want to pack in as many
rectangle as possible, given a sequence of requests.
Achieving either one is hard. |
|
x*******5
发帖数: 152 | 32 我暑假G实习，进行conversion process。10月初过第一个hiring committee。本来我
以为下面会很快。结果recruiter说现在无法提交到下一个vp/svp review，因为正在帮
我找组，然后3周过去了，没有update。请教各位googler，这个组的allocation
process如何，需要多长时间（in average). 我在glassdoor看见一个惊悚的例子，去
年11月面，组allocation花了4.5个月，今年5月才拿到offer。不知道自己是不是worst
case。有什么方法可以加快这个过程？谢谢大家 |
|
a*****1
发帖数: 314 | 33 why do we use circular link list in place of any balanced binary search tree
in storage allocator? One draw back is that to free() a chunk of memory
allocator has to search the link list and then if found that address then
release , so why not tree to reduce this search and merge?
请高手，给出简单解释。
谢谢 |
|
g*********e
发帖数: 14401 | 34 there are 4 conditions, all of which have to be satisfied for deadlock to
happen.
A deadlock situation can arise if and only if all of the following
conditions hold simultaneously in a system:[1]
Mutual Exclusion: At least one resource must be non-shareable.[1] Only one
process can use the resource at any given instant of time.
Hold and Wait or Resource Holding: A process is currently holding at least
one resource and requesting additional resources which are being held by
other processes.
No Pr... 阅读全帖 |
|
d*******u
发帖数: 186 | 35 From c programming language:
typedef long Align; /* for alignment to long boundary */
union header { /* block header */
struct {
union header *ptr; /* next block if on free list */
unsigned size; /* size of this block */
} s;
//The Align field is never used; it just forces each header to be
aligned on a worst-case boundary.
Align x; /* force alignment of blocks */
};
typedef union header Header;
static Header base; /* empty list to get started */
static Header *fr... 阅读全帖 |
|
s**x
发帖数: 7506 | 36
New head is just a pointer, when declare a pointer variable, only the memory
for the pointer is allocated, what the pointer points to is not allocated.
And in your case, this pointer value is not initialized , so it actually
points to a random object, so it is really bad. |
|
f****s
发帖数: 74 | 37 calxeda的电面就两个人，一人一题。
1. some_struct* p=NULL;
p->some_member;
问会发生什么？答：会编译通过，但运行会出问题，继续问为什么会出问题，怎么样出
问题，扯了os，trap，interrupt一堆。最后又问，没有os的imbeded system怎么处理
？瞎扯了boot loader，硬件interrupt handler什么的。面完上机试了一下，不会出错
，编译器优化了。
2.问memory alignment，如何allocate一块memory总是以64的整数倍地址开始。我就说
naive的方法是用malloc然后检查地址，他所不naive，问怎么样实现，我想了一会说必
须把原有的地址存下来，就多allocate一点把原有的地址存存下来，当用户调用free的
时候就用这个地址。感觉答的还不错，后来写code，出了个bug，经指点，纠正了。
今天收到hr的据信。感觉小公司就问些实际应用的东西，真不容易啊。 |
|
f****s
发帖数: 74 | 38 calxeda的电面就两个人，一人一题。
1. some_struct* p=NULL;
p->some_member;
问会发生什么？答：会编译通过，但运行会出问题，继续问为什么会出问题，怎么样出
问题，扯了os，trap，interrupt一堆。最后又问，没有os的imbeded system怎么处理
？瞎扯了boot loader，硬件interrupt handler什么的。面完上机试了一下，不会出错
，编译器优化了。
2.问memory alignment，如何allocate一块memory总是以64的整数倍地址开始。我就说
naive的方法是用malloc然后检查地址，他所不naive，问怎么样实现，我想了一会说必
须把原有的地址存下来，就多allocate一点把原有的地址存存下来，当用户调用free的
时候就用这个地址。感觉答的还不错，后来写code，出了个bug，经指点，纠正了。
今天收到hr的据信。感觉小公司就问些实际应用的东西，真不容易啊。 |
|
g********E
发帖数: 178 | 39 查了下原来析构是destruct，明白了，如果我就用默认的，
This destroys all container elements, and deallocates all the storage
capacity allocated by the vector using its allocator.
这样是不是析构就不如clear/resize快？
resize不错，非常符合我的需要:)
call |
|
r********9
发帖数: 1116 | 40 来自主题: JobHunting版 - BB店面经 declare from new only, means allocation from stack should be forbidden.
This can be achieved by making constructors private. In addition to that,
you should define a static wrapper function within which to invoke
constructor and return a dynamically allocated object.
class heapObject{
private:
heapObject(){}
public:
static heapObject* create(){ return new heapObject;}
};
int main(){
heapObject* sample=heapObject::create();
} |
|
J**9
发帖数: 835 | 41 1) For this signature
char* add(const char*a, const char *b)
It's definitely caller-free since you can't modify either one.
(allocation within the function)
(Or use a static buffer, but not recommended)
2) You can overwrite a with this, assuming a has enough spaces:
char* add(char*a, const char *b)
3) It's all caller's responsibility:
char* add(const char*a, const char *b, char *results)
Caller pre-allocates a buffer and passes in, of course, frees it later IF it
's from heap.
Choice 3) is the r... 阅读全帖 |
|
r*****e
发帖数: 792 | 42 我是不觉得，下标就老容易搞糊涂，还有有时候allocate vector时要多allocate
一个，这样就更容易出错了。总之，我是觉得DP比别的难些，水平有限啊。 |
|
c******n
发帖数: 4965 | 43 ft killed by their interview
most of the interviews I actually felt fine. 2 had bugs on first try, but
they point out and said there is a bug, I found it, and corrected, really
minor issues. and I even thought that such an interactive bug-finding
process is a plus.
looks that particularly some LAO3 ZHONG1 are extremely picky on us. as an
experienced GUI1 GONG1, I really don't see the point of them requiring such
bug-free code on first try, in real life you test it out anyway. some of the
intervi... 阅读全帖 |
|
l**t
发帖数: 6971 | 44
in
别的方面不评论，不过这个memory allocation的问题，如果你每天要allocate几个甚
至几十个billion string的话，还是很重要的。 |
|
|
m**********o
发帖数: 63 | 46 课程是：Multimedia Quality of Service Networking，要写survey paper。
以下为topic：
1.Traffic characteristics of multimedia (e.g., H.265 and H.264 SVC
encoded video/audio streams)and 3D video the relationships to the media
quality
2. Caching and object replication for multimedia content distribution (
including push vs. pull)
3.Session set-up/teardown protocols for multimedia streaming, e.g., SIP,
SAP, SCTP.
4. Control protocols for multimedia streaming, e.g., RTP/RTCP, RTSP.
5. Determining network ... 阅读全帖 |
|
i*****h
发帖数: 1534 | 47 【 以下文字转载自 SanFrancisco 讨论区 】
发信人: ninjaca (Ninja), 信区: SanFrancisco
标 题: 老印非法占中国绿卡名额, 大家去签名请愿. (转载)
发信站: BBS 未名空间站 (Sat Aug 17 02:57:14 2013, 美东)
发信人: ForgotMyPin (喵星人), 信区: JobHunting
标 题: 老印非法占中国绿卡名额, 大家去签名请愿. (转载)
发信站: BBS 未名空间站 (Fri Aug 16 21:40:59 2013, 美东)
大部分在美华人作为职业技术移民，却饱受等待绿卡的煎熬，最近华人在绿卡身份排期
上连吃暗亏。所谓爱哭的孩子有奶吃，自如果属于自己的正当权益都不去主动争取, 那
么最后也只有被别人任意欺凌的份了。希望我们这次能够做点什么，不管是签名，请愿
，打官司，我都支持，支持一切可以表达心声的活动。
下面是针对签证名额分配不公正待遇、数据公开等事项的向国会议员请愿书，希望大家
支持。
http://www.change.org/petitions/united-states-con... 阅读全帖 |
|
|
f******p
发帖数: 173 | 49 搜了一下，https://groups.google.com/forum/#!topic/microsoft.public.vc.mfc/
yBrUjgvGwf8 里面有一个回复我觉得挺不错，大家来说说怎样
Much better to allocate a block of storage (either a general "new" or
your own sub-allocator), copy the string into it, pass it to the recipient,
and let the
recipient release it (either delete or your own deallocation back to your
pool of
buffers). This means that there is never more than one thread, ever, using
the storage.
You do not protect the function; you protect the area it is copying ... 阅读全帖 |
|
w*********0
发帖数: 147 | 50 十月多电面了他家seattle office两面，然后就一直在wait.最近收到消息说他家给的
allocation今年严重缩水，导致我面的那个team没了allocation。就不继续forward了。
看来Intel被大Q挤的压力有点大。 |
|
